Rman 备份时RAC重启

Oracle 10gR2 RAC:

rman备份过程中数据库重启

由于安装数据的机器为p5 520性能较低，在运行rman备份过程中出现无响应等系统繁忙的症状，判断可能是系统响应不及时导致crs出现问题。检查crs日志/oracle/app/crs/log/db01/alertdb01.log，发现：

2012-01-15 01:06:02.715

[cssd(565310)]CRS-1601:CSSD Reconfiguration complete. Active nodes are db01 db02 .

2012-02-01 10:33:56.620

[cssd(565310)]CRS-1612:node db02 (0) at 50% heartbeat fatal, eviction in 0.000 seconds

2012-02-01 10:34:03.620

[cssd(565310)]CRS-1611:node db02 (0) at 75% heartbeat fatal, eviction in 0.000 seconds

2012-02-01 10:34:04.620

[cssd(565310)]CRS-1611:node db02 (0) at 75% heartbeat fatal, eviction in 0.000 seconds

2012-02-01 10:34:08.620

[cssd(565310)]CRS-1610:node db02 (0) at 90% heartbeat fatal, eviction in 0.000 seconds

2012-02-01 10:34:09.620

[cssd(565310)]CRS-1610:node db02 (0) at 90% heartbeat fatal, eviction in 0.000 seconds

2012-02-01 10:34:10.622

[cssd(565310)]CRS-1610:node db02 (0) at 90% heartbeat fatal, eviction in 0.000 seconds

2012-02-01 10:34:11.122

[cssd(565310)]CRS-1607:CSSD evicting node db02. Details in /oracle/app/crs/log/db01/cssd/ocs

sd.log.

可以判断是由于心跳无法传送导致。

所以寻求方法降低rman在备份时对系统的压力。

更改备份脚本如下：

db02$more rman.input

CONFIGURE RETENTION POLICY TO RECOVERY WINDOW OF 0 DAYS;

CONFIGURE CHANNEL DEVICE TYPE DISK RATE 8M;

run {
allocate channel d1 device type disk format '/backup/ora_rman/bck1%U';

sql 'alter system archive log current';
backup duration 00:15 minimize load archivelog all  delete all input;
backup duration 01:00 minimize load database;

crosscheck backup;
crosscheck archivelog all;
delete noprompt expired archivelog all;
delete noprompt expired backup;
delete noprompt obsolete;
release channel d1;
}

 

这样，将备份的时间延长到40分钟，（通常只需要5-10分钟即可备份完成），将磁盘速度限制在8M，从而达到降低rman备份负载的目的。